k0ntext

k0ntext provides a robust context engineering framework designed to bridge the gap between initial research and production code implementation. The Plan Phase skill specifically processes research manifests into actionable, chunk-based todolists. By identifying individual code chunks, their file-line references, and their dependency graphs, this tool ensures that each implementation step remains atomic, testable, and verifiable. It is built for senior developers and AI-assisted workflows using tools like Claude, Cline, Cursor, and Aider, ensuring that complex features are decomposed into manageable, high-success-rate execution units. The agent facilitates an Inter-Phase Awareness protocol, allowing the research phase to inform the implementation phase directly, reducing ambiguity and preventing common context-loss pitfalls during development.

• Automatically parses and validates research documents located in .claude/research/active/.

• Generates structured implementation manifests with atomic action items, risk assessments, and verification tests.

• Builds formal dependency graphs to enforce execution ordering (e.g., CHUNK-P1 → CHUNK-P2).

• Creates inter-phase contracts for automated consumption by the rpi-implement skill.

• Includes support for per-chunk rollback strategies, ensuring safe deployment cycles.

• Maintains a strict token budget for efficient context management during the planning process.

• Ensure a valid research document exists at .claude/research/active/[feature]_research.md before invocation.

• Use the provided /context-eng:research command if the research document is missing or outdated.

• Post-generation, proceed directly to the /context-eng:implement phase to execute the defined todolists.

• Monitor the chunk manifest status within the generated plan to track progress across the dependency graph.

• Use the agent to sync codebase knowledge, detect drifts, and validate documentation accuracy against current implementation state.

• The tool is optimized for Node.js 18+ environments and supports MCP (Model Context Protocol) for seamless integration with IDE-based AI agents.